Statistical Consulting Group - ETH Z · 2019-06-28 · Statistical Consulting Group Seminar for Statistics, ETH Z urich Seminar for Statistics, ETH Z urich (SfS) Introduction to ggplot2

Introduction to ggplot2

Statistical Consulting Group

Seminar for Statistics, ETH Zurich

Seminar for Statistics, ETH Zurich (SfS) Introduction to ggplot2 1 / 40

Table of Contents

1 Getting started with ggplot2

2 Aesthetics

3 Layers

4 Faceting

5 Tailoring Graphics


Getting Started with ggplot

In this section we will . . .

. . . get started with ggplot2

. . . create plots of one variable

. . . create plots of two variables

. . . learn how to save a plot


Must Have

Useful cheatsheet: https://www.rstudio.com/resources/cheatsheets/ (pickData Visualisation with ggplot2)

Source: link above. This image is under Creative Commons licence.


https://www.rstudio.com/resources/cheatsheets/

https://creativecommons.org/licenses/by/4.0/

Book Recommendations: R Graphics & ggplot

R Graphics CookbookWinston Chang, O’Reilly Media, 2012

and its online companion:http://www.cookbook-r.com/Graphs/

ggplot2: Elegant Graphics for Data Analysis (Use R!)Hadley Wickham, Springer, 2009

See also: https://ggplot2.tidyverse.org/


http://www.cookbook-r.com/Graphs/

https://ggplot2.tidyverse.org/

Why ggplot2?

Some advantages:

nice labels

nice colors

small margins

beautiful faceting or multipanel plots

very powerful and flexible: we will have a glimpse at the grammar ofgraphics

can easily change or update plots


Why ggplot2?

Some disadvantages:

ggplot2 can only deal with data.frames

default plots of model outputs are usually not possible

ggplot2 is not optimized for speed performance

3D plots are not possible


Functions in Package ggplot2

There are two important functions:

qplot: similar to base plotting functions (“for beginners”)

ggplot: the feature-rich “workhorse” (our focus)


Grammar of Graphics

The ”gg” in ggplot2 stands for grammar of graphics which is based onWilkinson’s (2005) grammar of graphics.

The grammar is useful because . . .

it is a generic way of creating a plot

it does not rely on a specific or customized graphic for a particularproblem

it allows for iterative updates of a plot

it uses the concepts of layers


Grammar of Graphics

Idea: all plots can be built from the same components

data set

coordinate system

aesthetic mapping that describes how information in data is beingmapped to visual properties (aesthetics) of geometric objects, socalled geoms.


Grammar of Graphics

Source: https://www.rstudio.com/resources/cheatsheets/



Grammar of Graphics

Source: https://www.rstudio.com/resources/cheatsheets/



Overview: Plots of One Variable

One continuous variable:

histogram: geom_histogram()

densities: geom_density()

frequency plot: geom_freqpoly()

One discrete (categorical) variable:

barplot: geom_bar()

pie plot: different coordinate system of barplot . . .


Illustration with mpg Data Set

Let us use the mpg data set from ggplot2.

It contains 234 observations about the fuel efficiency of 38 popular cars in1999 and 2008.

Let’s have a look at the democode.


Overview: Plots of Two Variables

Two continuous variables

scatter plot: geom_point()

scatter plot using jitter: geom_jitter()

smoother: geom_smooth()

Discrete x and continuous y

boxplot: geom_boxplot()

bar plot: geom_bar(stat = "identity")

Continuous function like time series

line plot: geom_line()


Illustration with mpg Data Set

Let’s look at the democode.


How to Save a Plot?

First we create an R object with the corresponding plot:> v <- ggplot(data = mpg, aes(x = class, y = cty)) + geom_boxplot() +

+ geom_jitter(alpha = 0.3)

Plots can then be saved by ggsave():> ggsave(filename = "cool-boxplot-II.png", plot = v)

ggsave automatically recognizes the output format (pdf, png, jpg, eps,svg)!


How to Save a Plot?

Control the width & height and change the path:> ggsave(filename = "cool-boxplot-III.jpg", plot = v, width = 5, height = 4,

+ path = "/path/of/figures/")

Alternatively, don’t forget to print the plot:> pdf("cool-boxplot-IV.pdf")

> print(v)

> dev.off()


Aesthetics

In this section we will have a look at the aesthetics . . .

. . . size

. . . shape

. . . color

. . . and combine them


Aesthetics: size

> ggplot(data = mpg, aes(x = hwy, y = cty, size = displ)) +

+ geom_jitter()

> # displ: engine displacement, in litres

●

●

●●

●

●●●

●

●●

●

●●

●●

●

●

●

●

●●●

●●

●●●

●

●●

●

●

●

● ●●

●

●●●

●●

●

●●

●●

●●

●

●●●

●

●●

●●

●

●

●

●

●●

●

●●●

●

● ●

●

● ●●●

●●

●●●●

●●●● ●

●

●

●●

●●

● ● ●●●

●

●

●

●

●

●

●

●

●

● ●

● ●

●●

●●

●●●

●

●

●●

●●●

●

●●

● ●●●● ●●●

●

●●●

●

●

● ●

●●●

●●

●●

●

●

●

●

●●

●

●●

●

●

●

●

●

●●

●

●●

●

●

●●

●● ●

●

●● ●

●

●●●

● ●●

●

●●●

●●

●

●

●

●

●

●

● ●

● ●● ●

●

●

●

●

●

●

●

●

●

●●●

●●

●

●

●

●

●●

●

●

●

●

●

●●

10

20

30

20 30 40hwy

cty

displ

●

●

●

●●●

2

3

4

5

6

7


Aesthetics: shape

> ggplot(data = mpg, aes(x = hwy, y = cty, shape = factor(cyl))) +

+ geom_jitter(size = 3)

> # we can set a fixed size for all the points

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●●

●

●●

●●

● ●

●●

●

●

● ●

●●

●●

●

●

●

●●●

● ●

●

●

●●

●● ●●

● ● ●●

● ●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

10

20

30

20 30 40hwy

cty

factor(cyl)

● 4

5

6

8


Aesthetics: color

> ggplot(data = mpg, aes(x = hwy, y = cty, color = factor(cyl))) +

+ geom_jitter(size = 3)

●

●●

●

●

● ●●

●

●

●

●

●●

●●

●

●

●

●

●●

●

●

●

●

●●

●

● ●

●

●

●

● ●

●

●

●

●●

●●

●

●●

●●

●

●

●

● ●●

●

●●

●●

●

●

●

●

●

●

●

●●

●

●

● ●

●

●●●

●

●

●

●

●●

●

●●

●● ●

●

●

●●

●●

● ● ●●

●

●

●●

●

●

●●

●

●

● ●

●●

●●

●● ●●

●

●●

●●

● ●●

●

●

●

●●

●●

● ●●

●

●

●●●

●

●

● ●

●●

●

●●

●

●

●

●

●

●●

●

●

●●

●●

●

●

●

●●

●

●●

●●

●●

●●

●

●

●● ●●

●●

●

●●

●

●

●● ●

● ●

●

●

●

●

●

●●

●

● ●●●

●

●

●

●

●

●

●

●

●

●●●

●

●

●

●

●

● ●●

●

●●

●

●

●

●

10

20

30

20 30 40hwy

cty

factor(cyl)

●

●

●

●

4

5

6

8


Aesthetics: Combination

> ggplot(data = mpg, aes(x = hwy, y = cty, color = factor(cyl),

+ shape = factor(cyl), size = displ)) +

+ geom_jitter()

> # there is only one combined legend for shape and color

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

● ●

●●

● ●

●●●

●

●●

●●

●

●

●

●

●

●●●

● ●●

●

●

●

●● ●●●

● ●●

● ●

●

●

●

●●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

10

20

30

20 30 40hwy

cty

displ

●

●

●

●●●

2

3

4

5

6

7

factor(cyl)

● 4

5

6

8


Aesthetics: Setting vs. Mapping

> ggplot(data = mpg, aes(x = hwy, y = cty, color = "blue", size = 2)) +

+ geom_jitter()

●

● ●●

●●●●

●

●●

●

●●

●●

●●

●

●

●●●

●●

●●●●

● ●

●

●

●

● ●●

●●

●●●●

●

●● ●●●●●

● ●●

●●●●●

●

●

●●

● ●

●

●●●

●

● ●

●

● ●●●

● ●● ●●●●●●● ●●

●

●●●

●● ● ●●

●

●

●●●

●

●●●

●

●●

● ●

●●●● ●●●

●●

●●

● ●●

●

●●

● ● ●●● ●●●

● ●●●

●●

●●

●●●●●

●●

●

●

●●● ●

●●●

●●

●

●

●●●●

● ●●

●

●●

●●●

●

●● ●●

●●●

● ● ●●

●● ●

● ●

●●

●

●

●

●●

●

● ●●●

●

●●●

●

●

●●

●●●●

●●

●

●

●

●● ●

●

●●

●

●

●●

10

15

20

25

30

35

10 20 30 40hwy

cty

colour

● blue

size

● 2

The color argument in aes will create a new variable with a single entry"blue" that is mapped to color (getting the first default color).Similarly for size.

In addition, a legend is being created.


Aesthetics: Setting vs. Mapping

If we want to set color to the explicit value "blue", we can do this in thecorresponding layer (outside of aes). Similarly for size.

> ggplot(data = mpg, aes(x = hwy, y = cty)) +

+ geom_jitter(color = "blue", size = 2)

●

●

●●

●

● ●●

●

●●

●

●●

●●

●●

●

●

●●

●

●● ●

●●

●

● ●

●

●

●

● ●

●●●

● ●

●●

●

●●●●

●

●●

● ●●

●

●●

●●

●

●

●

●

●

●

●

●●

●

●

● ●

●

● ●●

●

● ●●

●●●

●●

●● ●

●

●

●●

●

●● ● ●●

●

●

●

●

●●

●

●●

●

● ●

●●

●●●

●●

●●

●●

●●

● ●

●

●

●

●

● ●

●●

●●●

●

●

●●●

●

●

● ●

●●●

● ●

●

●

●

●

●

●●

●

●

●●

●

●

●

●

●

●●●

● ●

●

●

●

●

●●

●

●

●● ●●

●●

●

● ● ●●

●● ●

● ●

●

●

●

●

●

●

●●

●●

●

●

●

●

●

●

●

●

●

●

●● ●

●

●

●

●

●

●

●

● ●●

●

●

●

●

●●

10

15

20

25

30

35

10 20 30 40hwy

cty


Layers

In this section we will look at. . .

. . . where to place which arguments.

. . . the order of layers.


Layers: Where to Place Which Arguments?

All arguments specified in function ggplot() are passed to allsubsequent layers.

This holds true unless a layer contains another specification.

Arguments specified in a single layer only affect the correspondinglayer.



Basic plot to start with> ggplot(data = mpg, aes(x = hwy, y = cty)) + geom_jitter() + geom_smooth()

Color both points and smoothers per group of drv⇒ three smoothers are fitted:

> ggplot(data = mpg, aes(x = hwy, y = cty, color = drv)) + geom_jitter() +

+ geom_smooth()

●

●

●

●

●

● ●●

●

●

●

●

●●

●●

●

●

●

●

●●

●

●

●

●●●

●

● ●

●

●

●

● ●

●

●

●

●●

●●

●

●●

●●

●●

●●

●●

●

●●

●●

●

●

●

●

●

●

●

●●

●

●

● ●

●

● ●●

●

●

●

●

●●●

●●●● ●

●

●

●●

●

●

● ● ●●●

●

●●

●

●

●

●●

●

● ●

● ●

●●

●● ●

●●

●

●

●●

● ●

●

●

●

●

● ●

●●

● ●●

●

●

●●●

●

●

● ●

●●

●●

●

●●

●

●

●

●

●

●

●

●●

●

●

●

●

●

● ●●

● ●

●

●

●●

●●

●

●

●●●●

●●

●

● ● ●

●

●●

●

● ●

●

●

●

●

●

●

●

●

● ●●

●

●

●

●●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

10

20

30

20 30 40hwy

cty

drv

●

●

●

4

f

r



Color the points per group of drv⇒ one smoother is fitted:

> ggplot(data = mpg, aes(x = hwy, y = cty)) + geom_jitter(aes(color = drv)) +

+ geom_smooth()

●

● ●●

●

●●●

●

●

●

●

●●

●●

●

●

●

●

●

●●

●

●

●

●●

●

● ●

●

●

●

●●

●●

●● ●

●●

●

●●

● ●

●

●

●

●●●

●

●●

●●

●

●

●

●

● ●

●

●●

●

●

●●

●

● ●●

●

●

●

●

●●●

●●

●●

●

●

●

●●

●

●

●● ●●

●

●

●

●

●

●

●

●●

●

●●

● ●

●●●●

●●●

●●

●●

●●

●

●

●●

●●

●●●

●●●

●

●●●

●

●

● ●

●●●

●●

●

●

●

●

●

●

● ●

●

●●

●

●

●

●

●

●

●●

●●

●

●

●

●

●●●

●

●●●

●

●●

●

● ● ●

●

● ● ●

● ●

●

●

●

●

●

●

●

●

● ●●

●

●

●

●

●

●

●

●

●

●

●●●

●

●

●

●

●

●

● ●

●

●●

●

●

●

●

10

20

30

20 30 40hwy

cty

drv

●

●

●

4

f

r



Color the smoothers per group drv

⇒ three smoothers are fitted:

> ggplot(data = mpg, aes(x = hwy, y = cty)) + geom_jitter() +

+ geom_smooth(aes(color = drv))

●

●

●●

●

●●

●

●

●

●

●

●●

●●

●

●

●

●

●

●●

●

●

●●

●

●

● ●

●

●

●

●●

●●

●

●●

●●

●

●●

●●●

●

●

●● ●

●

●●

●●

●

●

●

●

●

●

●

●●

●

●

● ●

●

●●

●

●

●

●

●

● ●●

●●

●● ●

●

●

●● ●

●●

● ●●

●

●

●

●

●●

●●

●

●

● ●

● ●

●●

●●

●●●

●●

●●

● ●

●

●

●●

● ● ●●● ●●●

●

●●●

●

●

● ●

●●

●

●●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●●●

● ●

●●

●

●

●●

●

●

●●

●●

●●

●

●●

●

●

●● ●

● ●

●

●

●

●

●

●

●

●

● ●●●

●

●

●

●

●

●

●

●

●

●●●

●

●

●

●

●

●● ●

●

●

●

●

●

●

●

10

20

30

20 30 40hwy

cty

drv

4

f

r


Layers: Order of Layers

Plot the points first and then add the layer with boxplots:

> ggplot(data = mpg, aes(x = class, y = cty)) +

+ geom_jitter(alpha = 0.4) +

+ geom_boxplot(aes(fill = class))

●

●

●

●

●

●

●

●

●

10

20

30

2seater compact midsize minivan pickup subcompact suvclass

cty

class

2seater

compact

midsize

minivan

pickup

subcompact

suv


Layers: Order of Layers

Plot the boxplot first and afterwards add the layer of points:


+ geom_boxplot(aes(fill = class)) +

+ geom_jitter(alpha = 0.4)

●

●

●

●

●

●

●

●

●

10

20

30

2seater compact midsize minivan pickup subcompact suvclass

cty

class

2seater

compact

midsize

minivan

pickup

subcompact

suv


Faceting

In this section we will . . .

. . . consider faceting or multi-panel conditioning plots


Faceting: facet_wrap

Let’s look a multi-panel plots> # subset of the mpg data set

> mpg.small <- subset(mpg, manufacturer %in%

+ c("ford", "land rover", "toyota",

+ "chevrolet", "honda"))#, "volkswagen"))

> ggplot(data = mpg.small, aes(x = hwy, y = cty)) +

+ geom_jitter() + facet_wrap(~ manufacturer)

●

●

●●

●

●●

●●

●●

● ●

●

●

●

● ●●

●

●●●

●●●

●

●●

●●●●●

●●●

●

●

●●●

●●● ●●

●

●●

●●

●

●

●●

●●

●●●

● ● ●●

●● ●

● ●

●

●

●

●

●

●●

●

● ●●

●

●

●●

●●

●

●●

●

land rover toyota

chevrolet ford honda

20 30 20 30

20 3010

15

20

25

10

15

20

25

hwy

cty


Faceting: facet_grid

> ggplot(data = mpg.small, aes(x = hwy, y = cty)) +

+ geom_jitter() + facet_grid(manufacturer ~ cyl)

●

●

●

●●

● ●●

●●

●

●●

●● ●●● ● ●●

● ●●

●●

●● ●

● ●●

●●● ●

●●

●● ●●

●●●

●●●●● ●

● ●● ●

●

●

●●●

●●

●●●

●

● ●

●

●● ●●●●● ●

●●

● ●●●●

● ●●●

●

●●

4 6 8chevrolet

fordhonda

land rovertoyota

15 20 25 30 35 15 20 25 30 35 15 20 25 30 35

10

15

20

25

10

15

20

25

10

15

20

25

10

15

20

25

10

15

20

25

hwy

cty


Changing Colors or the Theme

In this section we will look at . . .

. . . how to select colors

. . . themes


How to Select Colors?

> require(RColorBrewer)

> display.brewer.all()

BrBGPiYG

PRGnPuOrRdBuRdGyRdYlBu

RdYlGnSpectral

AccentDark2Paired

Pastel1Pastel2

Set1Set2Set3

BluesBuGnBuPuGnBu

GreensGreys

OrangesOrRdPuBu

PuBuGnPuRd

PurplesRdPuRedsYlGn

YlGnBuYlOrBrYlOrRd



> # use brewer

> ggplot(data = mpg, aes(x = hwy, fill = drv)) + geom_density(alpha = 0.5) +

+ scale_fill_brewer(palette = "Dark2")

0.00

0.05

0.10

20 30 40hwy

dens

ity

drv

4

f

r



Websites helping you to select colorshttp://colorbrewer2.org/http://tools.medialab.sciences-po.fr/iwanthue/

Define your own colors> ggplot(data = mpg, aes(x = hwy, fill = drv)) +

+ geom_density(alpha = 0.5) +

+ scale_fill_manual(values = c("red", "green", "black"))


http://colorbrewer2.org/

http://tools.medialab.sciences-po.fr/iwanthue/

Themes

Change the theme of a plot using theme_...().

Let’s have a look at the democode.


+ geom_boxplot(aes(fill = class)) +

+ geom_jitter(alpha = 0.4) +

+ theme_bw()

See the R package ggthemes for additional themes.


Documents

Statistical Consulting Group - ETH Z · 2019-06-28 · Statistical Consulting Group Seminar for Statistics, ETH Z urich Seminar for Statistics, ETH Z urich (SfS) Introduction to ggplot2